Single- versus multi-tenant cloud architectures

shutterstock_519970543 (1)A few weeks back the topic of single-tenancy versus multi-tenancy in cloud architectures was mentioned. I made the point that single-tenancy has a major advantage over multi-tenancy - a perception by users that their data is more secure.

Arguing for single-tenancy is not the populist position. The approach is critiqued as "a SaaS-querade". It's often derided as another revival of the old ASP model, trotted out by traditional software vendors who are too risk-adverse to 'do' the cloud properly. They are content to sweat their on-premise legacy software asset one more way by adopting a "cloud strategy" that's really nothing of the sort.

As I said then: Like most things in IT, opinions differ.

So I wanted to re-examine the arguments pro the multi-tenancy model; to see why it is considered the 'true path' for cloud computing.

One view is that vendors using single, per-customer instances miss the "economies of scale" opportunity and that's extremely adverse for revenue growth and scalability.

That may be a true argument, but it isn't an especially customer-centric one.

Another argument is that only multi-tenancy provides the degree of data analytics to make it worthwhile for monetization. Aggregating network effect data from multiple tenancies provides insights into user behaviour - single tenancy or on-premise does not.

Again that may be true, but it isn't putting the customer or integrity of their data before the interests of the cloud vendor.

Some of the arguments allegedly in favour of multi-tenancy seem to be quite the reverse. One article recently argued that the way forward for multi-tenant systems was to provide more "elasticity" when it came to scheduling upgrades, i.e. allowing companies to stay on an older release for longer rather than mass-upgrading them to the latest one. Apart from the fact this is a non-standard (and not helpful) use of the term, surely making this type of "elasticity" a differentiating business benefit is a facet of the single- rather than the multi-tenant model?

The problem with the argument that single-tenancy advocates are all risk-adverse legacy software vendors is that it is illogical. It only takes one "born for the cloud" software vendor to opt for a single-tenancy model in order to debunk that argument.

One pro multi-tenancy blog writes: "It's always about your customer... Cloud or no Cloud." A good point, and well made, but where then is the application of that in using arguments that are all about the revenue of the cloud software vendor?

This debate is based on an unreliable foundation anyway. It's all multi-tenant at the infrastructure level where services such as fault tolerance/failover, load balancing, etc. are concerned. People talk about the degree of multi-tenancy and for IaaS (Infrastructure as a Service), PaaS (Platform as a Service) and SaaS (Software as a Service) it probably means something different for each.

The keywords for IaaS for end-customers tend to be words like secure, scalable, and resilient. None of those I'd argue are the preserve of multi-tenant deployment, and maybe even the opposite is true. Any application that needs its own secure and exclusive virtual computing environment where system performance is not affected by a 'noisy neighbour' would seem to be better-served by the single-tenancy model.

But there's no business model in it, say the pro multi camp. Surely that's down to the old-school exercise of making your revenue exceed your costs? If you can give customers more of what they want with the single-tenant model, then not only is it a route to profitability, it may even be a differentiator.

Value of a DMS for product development

Tags: Cloud, Data security, Private Cloud